Method, device, electronic device and storage medium for determining severity of coronary artery stenosis

Abstract

The present disclosure provides a method, device, electronic device and storage medium for determining the severity of coronary stenosis, the method comprising: obtaining actual blood lipid parameters including low-density lipoprotein cholesterol of a target patient; obtaining clinical characteristics including identity information and past medical history of the target patient; using the actual blood lipid parameters and clinical characteristics as input information, inputting a coronary stenosis severity prediction model obtained through pre-training, the coronary stenosis severity prediction model adopts a model framework of an interpretable enhancement machine; receiving the actual coronary stenosis score and score explanation information output by the coronary stenosis severity prediction model; determining the actual coronary stenosis severity of the target patient based on the actual coronary stenosis score and score explanation information. The application of this method can more accurately determine the association between low-density lipoprotein cholesterol and the severity of coronary stenosis, and the result has stronger interpretability.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, specifically to the field of artificial intelligence technology such as machine learning, association prediction, and auxiliary discrimination, and more particularly to a method, device, electronic device, computer-readable storage medium, and computer program product for determining the severity of coronary artery stenosis. Background Art

[0002] In the field of cardiovascular disease, atherosclerosis (ASCVD) is the main pathological mechanism of coronary heart disease and acute myocardial infarction, and low-density lipoprotein cholesterol (LDL-C) is considered to be one of the core factors leading to atherosclerosis. Therefore, the control of LDL-C levels is regarded as a key link in the prevention and treatment of cardiovascular diseases. The European Society of Cardiology guidelines divide the control targets of LDL-C into multiple risk levels and propose strict LDL-C control targets for high-risk patients. For example, for very high-risk ASCVD patients, LDL-C levels should be controlled below 1.4 mmol / L.

[0003] Although a large number of studies have shown that LDL-C levels are closely related to atherosclerosis and coronary artery stenosis, and the concept of "the lower the LDL-C, the better" has been widely recognized, the existing clinical evaluation methods are insufficient in identifying the refined risks brought about by different LDL-C levels, especially the inability to capture the complex nonlinear relationship between LDL-C and coronary artery stenosis, making it difficult to develop more individualized treatment plans for different patients. Summary of the invention

[0004] The embodiments of the present disclosure provide a method, device, electronic device, computer-readable storage medium, and computer program product for determining the severity of coronary artery stenosis.

[0005] In a first aspect, the present disclosure provides a method for determining the severity of coronary artery stenosis, comprising:

[0006] Obtain the actual blood lipid parameters including low-density lipoprotein cholesterol of the target patient;

[0007] Obtain clinical characteristics of target patients including identification information and medical history;

[0008] The actual blood lipid parameters and clinical characteristics are used as input information and input into a pre-trained coronary artery stenosis severity prediction model; wherein the coronary artery stenosis severity prediction model is trained using training samples formed by sample inputs of actual blood lipid parameters and clinical characteristics of historical patients and sample outputs of corresponding true coronary artery stenosis scores, and the coronary artery stenosis severity prediction model adopts a model framework of an interpretable enhancement machine;

[0009] receiving an actual coronary artery stenosis score and score interpretation information output by a coronary artery stenosis severity prediction model;

[0010] Based on the actual coronary artery stenosis score and the score interpretation information, the actual severity of coronary artery stenosis in the target patient is determined.

[0011] Optionally, the actual blood lipid parameters include at least low-density lipoprotein cholesterol and at least one of high-density lipoprotein cholesterol, total cholesterol, and triglycerides.

[0012] Optional identity information includes: gender information, age information, and past medical history includes: whether there is a history of any of hypertension, diabetes, kidney disease, and stroke.

[0013] Optionally, before using the sample input and sample output to form a training sample, the following is further included:

[0014] Interpolate missing values ​​in sample input and sample output;

[0015] Outliers in sample input and sample output are removed or replaced so that the interpolated and removed or replaced data constitute training samples.

[0016] Optionally, interpolate missing values ​​in sample input and sample output, including:

[0017] For continuous variables with missing values ​​in sample input and sample output, median interpolation method is used for interpolation;

[0018] For categorical variables with missing values ​​in sample input and sample output, the majority interpolation method is used for interpolation.

[0019] Optionally, score explanation information includes:

[0020] At least one of the following: variable importance analysis results, global explanation information, interaction analysis information, and local explanation information;

[0021] Among them, the results of variable importance analysis are used to present the independent influence of different types of input variables on the actual coronary stenosis score; the global explanatory information is used to present the actual corresponding relationship between the variable value of each input variable and the coronary stenosis score; the interaction analysis is used to present the combined influence of different types of input variables on the actual coronary stenosis score when they are combined with each other; the local explanatory information is used to present the contribution polarity of the variable value of each input variable to the actual coronary stenosis score for the target patient, and the contribution polarity includes positive contribution of positive correlation and negative contribution of negative correlation.

[0022] Optionally, the score explanation information also includes: the risk level to which the actual coronary artery stenosis score belongs; wherein the risk threshold for dividing the risk levels is determined based on a coronary artery stenosis severity prediction model by using training samples.

[0023] In a second aspect, the present disclosure provides a device for determining the severity of coronary artery stenosis, comprising:

[0024] an actual blood lipid parameter acquisition unit, configured to acquire actual blood lipid parameters including low-density lipoprotein cholesterol of a target patient;

[0025] A clinical characteristic acquisition unit is configured to acquire clinical characteristics of a target patient including identity information and past medical history;

[0026] The actual parameter input unit is configured to input the actual blood lipid parameters and clinical characteristics as input information into a pre-trained coronary stenosis severity prediction model; wherein the coronary stenosis severity prediction model is trained using training samples formed by sample inputs of the actual blood lipid parameters and clinical characteristics of historical patients and sample outputs of the corresponding true coronary stenosis scores, and the coronary stenosis severity prediction model adopts a model framework of an interpretable enhancement machine;

[0027] A model output result receiving unit is configured to receive an actual coronary artery stenosis score and score interpretation information output by the coronary artery stenosis severity prediction model;

[0028] The actual coronary artery stenosis severity determination unit is configured to determine the actual coronary artery stenosis severity of the target patient based on the actual coronary artery stenosis score and the score interpretation information.

[0029] Optionally, the actual blood lipid parameters include at least low-density lipoprotein cholesterol and at least one of high-density lipoprotein cholesterol, total cholesterol, and triglycerides.

[0030] Optional identity information includes: gender information, age information, and past medical history includes: whether there is a history of any of hypertension, diabetes, kidney disease, and stroke.

[0031] Optionally, before using the sample input and sample output to form a training sample, the following is further included:

[0032] A missing value interpolation unit is configured to interpolate missing values ​​in sample input and sample output;

[0033] The outlier processing unit is configured to remove or replace outliers in sample input and sample output, so that the data after interpolation and removal or replacement constitutes training samples.

[0034] Optionally, the missing value imputation unit is further configured as:

[0035] For continuous variables with missing values ​​in sample input and sample output, median interpolation method is used for interpolation;

[0036] For categorical variables with missing values ​​in sample input and sample output, the majority interpolation method is used for interpolation.

[0037] Optionally, score explanation information includes:

[0038] At least one of the following: variable importance analysis results, global explanation information, interaction analysis information, and local explanation information;

[0039] Among them, the results of variable importance analysis are used to present the independent influence of different types of input variables on the actual coronary stenosis score; the global explanatory information is used to present the actual corresponding relationship between the variable value of each input variable and the coronary stenosis score; the interaction analysis is used to present the combined influence of different types of input variables on the actual coronary stenosis score when they are combined with each other; the local explanatory information is used to present the contribution polarity of the variable value of each input variable to the actual coronary stenosis score for the target patient, and the contribution polarity includes positive contribution of positive correlation and negative contribution of negative correlation.

[0040] Optionally, the score explanation information also includes: the risk level to which the actual coronary artery stenosis score belongs; wherein the risk threshold for dividing the risk levels is determined based on a coronary artery stenosis severity prediction model by using training samples.

[0041] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that when the at least one processor executes, it is possible to implement the method for determining the severity of coronary stenosis as described in any implementation method in the first aspect.

[0042] In a fourth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, which are used to enable a computer to implement the method for determining the severity of coronary stenosis as described in any implementation method of the first aspect when executed.

[0043] In a fifth aspect, an embodiment of the present disclosure provides a computer program product comprising a computer program, which, when executed by a processor, can implement the steps of the method for determining the severity of coronary artery stenosis as described in any implementation manner in the first aspect.

[0044] The scheme for determining the severity of coronary artery stenosis provided by the present invention adopts an interpretable boosting machine model that expands the regression tree-based boosting method on a generalized additive model as a model framework, and adopts blood lipid parameters including low-density lipoprotein cholesterol and clinical characteristics including identity information and past medical history as input variables, so that multiple types of influencing features can be used to jointly predict the accurate relationship between the severity of coronary artery stenosis. Moreover, since the above-mentioned multivariate historical parameters are used as training samples, the characteristics of the interpretable boosting machine model can provide clear feature explanations and interaction effects between variables while ensuring prediction accuracy, so that the model can also output explanatory information on how to obtain the results, thereby enhancing doctors' recognition of the output results and better assisting doctors in making more accurate diagnoses.

[0045] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Other features, objects and advantages of the present disclosure will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0047] Figure 1 A flowchart of a method for determining the severity of coronary artery stenosis provided in an embodiment of the present disclosure;

[0048] Figure 2 A flowchart of a method for preprocessing sample input and sample output provided in an embodiment of the present disclosure;

[0049] Figure 3 A schematic diagram of the composition structure of score explanation information provided by an embodiment of the present disclosure;

[0050] Figure 4-1 A schematic diagram showing a global explanation of an embodiment of the present disclosure;

[0051] Figure 4-2A schematic diagram of an overall explanation of the predictive relationship between LDL-C and Gensini score shown in an embodiment of the present disclosure;

[0052] Figure 4-3 A schematic diagram of an overall explanation of the predictive relationship between an interaction term (age and LDL-C) and the Gensini score according to an embodiment of the present disclosure;

[0053] Figure 4-4 A schematic diagram showing a partial explanation of an embodiment of the present disclosure;

[0054] Figure 5 A structural block diagram of a device for determining the severity of coronary artery stenosis provided in an embodiment of the present disclosure;

[0055] Figure 6 A schematic diagram of the structure of an electronic device suitable for executing a method for determining the severity of coronary stenosis provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0056] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. It should be noted that the embodiments in the present disclosure and the features in the embodiments may be combined with each other without conflict.

[0057] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0058] At present, many coronary artery disease risk assessment models are mainly based on statistical methods, such as linear regression models, logistic regression models or Cox proportional hazard models. These models analyze the relationship between LDL-C and coronary artery stenosis in a linear or approximately linear manner. Although this method is simple and easy to use, it has obvious limitations, namely, it cannot accurately capture complex nonlinear relationships. In addition, the linear regression-based method assumes that the relationship between variables is linear, and fails to fully reflect the nonlinear effect between LDL-C and atherosclerotic plaque formation and coronary artery stenosis. There are also some complex machine learning models such as deep neural networks (DNN) that have been applied to the field of coronary heart disease risk prediction, but the "black box" characteristics of these models make them difficult to use in clinical practice because doctors cannot clearly explain the basis for the model's predictions.

[0059] For clinical assessment of the severity of coronary artery stenosis, the degree of arterial stenosis is usually quantitatively analyzed by coronary angiography and calculation of the Gensini score. The Gensini score is a commonly used tool to assess the severity of coronary artery lesions, and the higher the score, the more severe the coronary artery stenosis. Although the Gensini score can well quantify the degree of coronary artery stenosis, it relies on complex and invasive clinical testing methods and cannot directly establish an accurate association with blood lipid indicators such as LDL-C.

[0060] Through the analysis of the above-mentioned related technologies and combined with the applicant's long-term clinical experience, it is believed that they have the following technical defects:

[0061] 1) Limitations of linear assumptions: Existing linear regression or logistic regression models assume that the relationship between LDL-C and the degree of coronary artery stenosis is linear. However, actual clinical data show that the relationship between LDL-C and coronary artery stenosis is often nonlinear, especially in different patient groups, affected by multiple factors such as age, gender, and previous medical history. LDL-C risk assessment requires a more flexible nonlinear model;

[0062] 2) Lack of ability to handle complex interactions: Existing models often ignore the complex interactions between LDL-C and other variables (such as age, gender, blood pressure, diabetes, etc.). These interactions play an important role in the pathological process of coronary heart disease, but traditional linear models are difficult to capture these details;

[0063] 3) Poor model interpretability: Although deep learning and other complex machine learning models have high prediction accuracy, their "black box" characteristics make the prediction results of these models difficult to explain, resulting in doctors being unable to trust the output of these models in clinical practice. Doctors often need to have a clear understanding of the basis for risk prediction in order to accurately apply it to patient management and decision making in clinical practice.

[0064] The execution subject of the method for determining the severity of coronary stenosis provided by the present disclosure may be a terminal device, a computer, or a server, or other devices with data processing capabilities. There is no limitation on the execution subject of the method.

[0065] Optionally, the terminal device can be a mobile phone, or a tablet computer, a wearable device, a vehicle-mounted device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc. The embodiments of the present disclosure do not limit the specific type of the terminal device.

[0066] In some embodiments, the server may be a single server, or a server cluster composed of multiple servers. In some implementations, the server cluster may also be a distributed cluster. The present disclosure does not limit the specific implementation of the server.

[0067] Please refer to Figure 1 , Figure 1 A flowchart of a method for determining the severity of coronary artery stenosis provided in an embodiment of the present disclosure, wherein process 100 includes the following steps:

[0068] Step 101: Acquire actual blood lipid parameters including low-density lipoprotein cholesterol of a target patient;

[0069] This step is intended to obtain the actual blood lipid parameters of the target patient including low-density lipoprotein cholesterol by the execution subject of the method for determining the severity of coronary artery stenosis (such as a terminal device or server with data analysis and processing capabilities). That is, the actual blood lipid parameters should at least include: low-density lipoprotein cholesterol, and in addition, may also include: at least one of high-density lipoprotein cholesterol, total cholesterol, and triglycerides.

[0070] The patient's actual blood lipid parameters are important parameters for assessing cardiovascular health and metabolic status. Specifically, actual blood lipid parameters include low-density lipoprotein cholesterol (LDL-C) and other key lipid indicators. The following is a detailed description of these indicators:

[0071] Low-density lipoprotein cholesterol (LDL-C): It is a type of lipoprotein in the blood that is mainly responsible for transporting cholesterol from the liver to various tissues in the body. High levels of LDL-C are associated with an increased risk of cardiovascular diseases such as atherosclerosis and heart disease. It is usually obtained through blood tests, and the common unit is milligrams per deciliter (mg / dL).

[0072] High-density lipoprotein cholesterol (HDL-C): Also known as "good cholesterol" because it helps recycle cholesterol from the blood and tissues and transport it back to the liver, reducing the risk of heart disease. Higher HDL-C levels are generally associated with a lower risk of cardiovascular disease.

[0073] Total cholesterol (TC): It is the sum of all cholesterol in the blood, including LDL, HDL and other types of cholesterol. The total cholesterol level is usually checked to assess the overall risk of cardiovascular disease.

[0074] Triglyceride (TG): It is a type of fat in the blood that is mainly used to store energy. High levels of triglycerides are also associated with an increased risk of cardiovascular disease, and maintaining normal triglyceride levels is essential for preventing metabolic syndrome and heart disease.

[0075] In clinical practice, obtaining the above blood lipid parameters can help doctors better perform the following operations: 1) Risk assessment: understand the patient's cardiovascular disease risk and guide lifestyle changes and treatment plans; 2) Treatment monitoring: adjust drug treatment (such as statins) according to blood lipid levels; 3) Personalized medicine: develop a personalized health management plan based on the patient's specific blood lipid conditions.

[0076] That is, this step obtains actual blood lipid parameters including LDL-C. On the basis of taking LDL-C as the core indicator, it also enriches the input variables by comprehensively considering multiple indicators, so as to more comprehensively clarify the patient's blood lipid parameters, thereby facilitating the model to learn more and deeper knowledge.

[0077] Step 102: Obtain clinical characteristics of the target patient including identity information and medical history;

[0078] Based on step 101, this step aims to obtain the clinical characteristics of the target patient including identity information and medical history by the above-mentioned execution subject. The identity information includes: gender information, age information, and the medical history includes: whether the patient has a history of any of hypertension, diabetes, kidney disease, and stroke.

[0079] Among them, the patient's gender information usually includes male and female, which is crucial for medical decision-making, because the incidence and symptoms of certain diseases may differ between different genders, and gender differences may affect drug metabolism, disease risk and treatment response (for example, the manifestations and risk factors of cardiovascular disease in women and men may be different); and the patient's age refers to the patient's actual age, usually in years, which is an important factor in assessing disease risk. The incidence of many diseases (such as cardiovascular disease, cancer, etc.) is closely related to age. Age may also affect treatment options and prognosis assessment.

[0080] Hypertension refers to a state of persistently high blood pressure, usually with a diagnostic standard of 140 / 90 mmHg or above. Hypertension is an important risk factor for cardiovascular disease. Knowing whether a patient has a history of hypertension can help doctors assess the risk of heart disease, stroke and other diseases.

[0081] Diabetes is a metabolic disease that affects how the body uses glucose. It is mainly divided into type 1 and type 2 diabetes. People with diabetes have a significantly increased risk of cardiovascular disease, kidney disease, and neuropathy. Obtaining a patient's diabetes history helps develop a diet, exercise, and drug treatment plan.

[0082] Kidney disease refers to a disease in which kidney function gradually declines, which is usually assessed by glomerular filtration rate (GFR). Kidney disease is closely related to cardiovascular disease. Knowing whether a patient has this medical history can guide kidney function monitoring and corresponding treatment measures.

[0083] Stroke is an acute disease caused by affected blood supply to the brain. It is usually divided into ischemic and hemorrhagic stroke. A previous history of stroke will significantly increase the patient's risk of another stroke. Understanding this history can help doctors conduct comprehensive risk assessments and develop prevention strategies.

[0084] This step aims to obtain clinical characteristics including identity information and past medical history to better understand the patient's actual health status, that is, the impact of historical health status on current and future health status.

[0085] Step 103: using actual blood lipid parameters and clinical characteristics as input information and inputting them into a pre-trained coronary artery stenosis severity prediction model;

[0086] On the basis of step 102, this step aims at having the above-mentioned execution subject input the actual blood lipid parameters and clinical characteristics as input information into the pre-trained coronary artery stenosis severity prediction model.

[0087] Among them, the coronary artery stenosis severity prediction model is trained using training samples formed by sample input consisting of actual blood lipid parameters and clinical characteristics of historical patients and sample output consisting of corresponding true coronary artery stenosis scores (that is, each training sample is a sample pair formed by sample input and sample output). The coronary artery stenosis severity prediction model adopts the model framework of Explainable Boosting Machine (EBM).

[0088] Among them, EBM is essentially an interpretable machine learning model, a variant of the generalized additive model, which aims to provide model interpretability while maintaining high prediction accuracy. EBM combines the advantages of ensemble learning and interpretable modeling, and is suitable for in-depth analysis of the impact of features.

[0089] It has the following main features:

[0090] 1) Interpretability: EBM allows users to understand the impact of each feature on the final prediction by decomposing the model's prediction into the contribution of each feature. This interpretability is particularly important in fields that require transparent decision-making processes, such as medicine, finance, and law.

[0091] 2) Generalized Additive Models (GAMs): The prediction result is the weighted sum of features. The contribution of each feature is modeled by a non-parametric function, allowing non-linear relationships to be captured.

[0092] 3) Ensemble learning: Using enhancement techniques, similar to Gradient Boosting Trees, models are gradually added to optimize the loss function. Each step improves on the previous model, thereby improving the overall performance.

[0093] 4) Handling feature interactions: It can naturally handle the interactions between different features. By modeling feature interactions, users can more clearly understand which feature combinations have the greatest impact on the prediction results.

[0094] 5) Efficiency: The training process is usually faster than traditional deep learning models and performs well on small to medium-sized datasets.

[0095] The present disclosure specifically uses the model trained by EBM to help doctors understand how the model arrives at a diagnosis result. Of course, in addition to being applied to this scenario, it can also be applied to financial service scenarios to explain the decision-making basis of credit scoring and risk assessment models, and to user behavior analysis scenarios to analyze the impact of user behavior characteristics on marketing strategies. That is, EBM can not only provide high-precision prediction results, but also ensure the transparency of the decision-making process, and is suitable for various application scenarios that require understanding the reasons behind model predictions. By using EBM, data scientists and business analysts can communicate model results more effectively and improve user trust.

[0096] Before data analysis, all collected patient data (i.e., the sample input and sample output mentioned above) can also be comprehensively preprocessed to ensure the integrity and consistency of the data. This can mainly include the following two aspects of processing: interpolation of missing values ​​in sample input and sample output, and removal or replacement of abnormal values ​​in sample input and sample output, so that the interpolated and removed or replaced data will eventually form training samples for subsequent training, and the same applies to test sets if necessary.

[0097] Specifically, for the interpolation of missing values, the median interpolation method can be used for continuous variables (such as LDL-C, TC, etc.); for categorical variables (such as gender, medical history, etc.), the mode interpolation method can be used. This method minimizes the impact of data loss on model training;

[0098] For the removal or replacement of outliers, we can first use statistical methods such as boxplot to identify outliers in the data, and then remove or replace the abnormal points in blood lipid indicators and Gensini scores that are extremely deviated from the normal range to prevent these outliers from having a negative impact on the stability of the model (the above process corresponds to Figure 2 Schematic diagram shown).

[0099] Step 104: receiving the actual coronary artery stenosis score and score interpretation information output by the coronary artery stenosis severity prediction model;

[0100] On the basis of step 103, this step aims at receiving, by the above-mentioned execution subject, the actual coronary artery stenosis score and score interpretation information output by the coronary artery stenosis severity prediction model.

[0101] Specifically, the score can still be obtained based on the Gensini scoring algorithm, and the score explanation information is used to explain how to obtain the actual coronary artery stenosis score, which may include at least one of the variable importance analysis results, global explanation information, interaction analysis information, and local explanation information (corresponding to Figure 3 Schematic diagram shown).

[0102] Among them, the variable importance analysis results are used to present the independent influence of different types of input variables on the actual coronary stenosis score; the global explanatory information is used to present the actual corresponding relationship between the variable value of each input variable and the coronary stenosis score; the interactive analysis is used to present the combined influence of different types of input variables on the actual coronary stenosis score when combined with each other; the local explanatory information is used to present the contribution polarity of the variable value of each input variable to the actual coronary stenosis score for the target patient, and the contribution polarity includes positive contribution of positive correlation and negative contribution of negative correlation.

[0103] Furthermore, the score explanation information may also include: the risk level to which the actual coronary stenosis score belongs, such as low risk level, medium risk level, and high risk level, so as to correspond the severity and treatment methods through different risk levels, wherein the risk threshold for each risk level can be determined by using training samples based on the coronary stenosis severity prediction model.

[0104] Step 105: Determine the actual severity of the coronary artery stenosis of the target patient based on the actual coronary artery stenosis score and the score interpretation information.

[0105] Based on step 104, this step aims to enable the above-mentioned execution subject to push the actual coronary artery stenosis score and score interpretation information to the corresponding doctor as auxiliary diagnosis information, so that the doctor can make a final diagnosis and finally determine the actual severity of coronary artery stenosis of the target patient.

[0106] The method for determining the severity of coronary artery stenosis provided in the embodiment of the present disclosure adopts an interpretable booster model that extends a regression tree-based boosting method on a generalized additive model as a model framework, and adopts blood lipid parameters including low-density lipoprotein cholesterol and clinical characteristics including identity information and past medical history as input variables, so that multiple types of influencing features can be used to jointly predict the accurate relationship between the severity of coronary artery stenosis. Moreover, since the above-mentioned multivariate historical parameters are used as training samples, the characteristics of the interpretable booster model can provide clear feature explanations and interaction effects between variables while ensuring prediction accuracy, so that the model can also output explanatory information on how to obtain the results, thereby enhancing doctors' recognition of the output results and better assisting doctors in making more accurate diagnoses.

[0107] To deepen understanding, this disclosure also provides a set of specific implementation solutions in combination with a specific application scenario:

[0108] In view of the various technical defects existing in the existing related technologies, this embodiment proposes an LDL-C risk stratification scheme based on the EBM model, aiming to utilize the nonlinear regression capability and high interpretability of the model to accurately identify the impact of different levels of LDL-C on the risk of coronary artery stenosis and provide a more personalized and accurate risk assessment. At the same time, this embodiment hopes to provide clinicians with a reliable prediction basis through the transparency of the EBM model, so that it can be effectively applied in actual clinical decision-making.

[0109] In order to analyze the complex nonlinear relationship between LDL-C level and coronary artery stenosis (Gensini score), and to combine other cardiovascular risk factors (such as age, gender, medical history, etc.), it is necessary to first establish a machine learning model with good interpretability. This model has the following characteristics:

[0110] 1) Multivariate input: In addition to analyzing LDL-C levels, clinical characteristics including age, gender, and past medical history (such as diabetes, hypertension, and renal insufficiency) are also combined to comprehensively assess the patient's risk of coronary artery stenosis.

[0111] 2) Nonlinear modeling: The EBM model captures the nonlinear relationship between LDL-C and the severity of coronary artery stenosis through stepwise additive regression and accurately divides the risk threshold of LDL-C.

[0112] 3) High interpretability: The contribution and interaction of each feature output by the EBM model provide clinicians with a clear basis for prediction, making this method highly clinically practical.

[0113] The following will explain the main steps step by step:

[0114] Step 1: Data Collection

[0115] Based on a single-center cross-sectional study, the subjects were patients with acute myocardial infarction (AMI) who were treated in the Department of Cardiology at Beijing XX Hospital from September 2017 to June 2023. All patients were adults aged 18 years and above. Exclusion criteria included individuals with missing coronary angiography records and lipid test results. Patients were selected based on the diagnosis by a cardiologist and the degree of coronary artery stenosis confirmed by coronary angiography. The data source was the hospital's Electronic Medical Record (EMR) system and laboratory testing system.

[0116] The specific patient screening process is as follows: 1264 patients were retrieved from the electronic medical record system, and participants with missing coronary angiography records (n = 56) were first removed, and then participants with missing data on four blood lipid items [low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG)] (n = 8) were excluded, and finally a total of 1200 participants were obtained. Table 1 below shows the characteristics of the 1200 participants, and the values ​​in Table 1 are shown as medians [Q1, Q3] or n (%):

[0117] Table 1 Distribution of actual lipid parameters and clinical characteristics of 1200 participants

[0118]

[0119] The data in Table 1 include:

[0120] 1) Coronary angiography data: The degree of coronary artery stenosis of each patient was obtained through coronary angiography, and the Gensini score was used to quantitatively evaluate the degree of lesions. The Gensini score is scored based on the degree of stenosis of each coronary artery segment (expressed as a percentage) and the location of the lesion (different segments have different weights). The higher the score, the more severe the coronary artery stenosis.

[0121] 2) Blood lipid index: Through laboratory testing, key blood lipid data such as low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), triglycerides (TG) and other key blood lipid data of each patient are obtained. Among them, LDL-C is the core index of the present invention and is used to evaluate the main risk factors of coronary artery stenosis.

[0122] 3) Patient characteristics:

[0123] Age: The actual age of the patient, included in the model as a continuous variable.

[0124] Sex: male or female, as a binary categorical variable.

[0125] Medical history: including hypertension, diabetes, kidney disease, and previous stroke as binary categorical variables. These factors can affect the progression of coronary artery stenosis and are therefore important covariates in the model.

[0126] Step 2: Data Preprocessing

[0127] Before data analysis, this embodiment performs comprehensive data preprocessing on all collected patient data to ensure data integrity and consistency. The specific steps are as follows:

[0128] 1) Missing value processing: For missing variable values ​​in some patient data, a reasonable interpolation strategy is adopted. For continuous variables (such as LDL-C, TC, etc.), the median interpolation method is used; for categorical variables (such as gender, medical history, etc.), the mode interpolation method is used. This method minimizes the impact of data loss on model training.

[0129] 2) Outlier detection and processing: Statistical methods such as box plots are used to identify outliers in the data. Outliers that are extremely deviated from the normal range in blood lipid indicators and Gensini scores are removed or replaced to prevent these outliers from having a negative impact on the stability of the model.

[0130] Step 3: EBM model construction

[0131] The core of this embodiment is to model the relationship between LDL-C and the risk of coronary artery stenosis through the EBM model. The EBM model is an extension of the generalized additive model, which uses an enhanced method based on a regression tree to provide clear feature explanations and interaction effects between variables while ensuring prediction accuracy.

[0132] The steps to build the model are as follows:

[0133] 1) Dataset division: The data of 1,200 patients were randomly divided into a training set and a test set, with 70% of the data used for the training set and 30% for the test set. This division ensures that the model can learn data patterns on the training set and evaluate its generalization ability on the test set.

[0134] 2) Variable modeling: The regression EBM model was created using the "InterpretML" library of Python software. For each input variable (including LDL-C, age, gender, etc.), the EBM model was fitted stepwise using a regression tree. In each iteration, the model continuously updates the additive structure of each variable by minimizing the prediction error. Since there is a significant nonlinear relationship between LDL-C and coronary artery stenosis, the nonlinear regression capability of EBM is able to capture these complex patterns.

[0135] 3) Interaction Term Detection: The EBM model has the ability to automatically detect interactions. During model training, the model automatically identifies the interaction effects between LDL-C and other variables (such as age, gender, hypertension, etc.). For example, the model found that LDL-C has different effects on coronary artery stenosis in patients of different age groups, and this interactive relationship is difficult to capture in traditional linear models.

[0136] 4) Model training and parameter setting: The EBM model does not require feature scaling and is trained directly using default parameters. Each feature is fitted using a shape function, and the interactions between features are automatically detected and incorporated into the model based on actual data. Ultimately, the model's predictive ability is gradually optimized through iterative updates of the model.

[0137] Step 4: EBM model interpretation

[0138] The EBM model provides variable importance analysis, global explanation, interaction analysis, and local explanation, which can clearly show the contribution of each feature to the final prediction. The output of the model includes:

[0139] 1) Variable importance analysis: By calculating the average absolute score of each feature in the training set, this example ranks the importance of each variable, see Figure 4-1 . Among all the variables, age was found to be the most influential variable, followed by LDL-C and diabetes. The interpretability of the model allows doctors to clearly understand the impact of each variable on the risk of coronary stenosis. Here, the term "importance" refers to the average absolute contribution (score) of each term (feature or interaction) to the prediction. These contributions are averaged over the training dataset, taking into account the number of samples in each bin and the sample weights (if relevant)

[0140] 2) Global interpretation: Shows the relationship between each predictor variable and coronary artery stenosis (Gensini score). The score of each feature (y-axis) represents its contribution to the Gensini score. For example, the higher the LDL-C score, the more severe the predicted coronary artery stenosis. Figure 4-2 The nonlinear relationship between LDL-C and Gensini score is shown. It should be noted that because EBM is an additive model, Figure 4-2 The contribution of LDL-C to the Gensini score prediction can be precisely quantified. For example, if a new data point with LDL-C = 1mmol / L is input, the model will increase the final Gensini score prediction by about -7. The density is a histogram that depicts the data distribution of LDL-C. Error bars are represented by gray markers. These are approximations of the model's uncertainty about a particular feature region. The impact of a slight modification of the training data increases with the width of the bar.

[0141] 3) Interaction analysis: Figure 4-3 The predictive relationship between the interaction term (age and LDL-C) and the Gensini score was fully explained. When age > 75 years and LDL-C > 4.5mmol / L, the predicted Gensini score of the interaction term was > 3, indicating that when the values ​​of age and LDL-C were both higher than the corresponding range, the risk of coronary artery stenosis would increase significantly.

[0142] 4) Local explanation: For the prediction of a specific patient, the model can explain the specific contribution of each feature to the Gensini score of that patient, see Figure 4-4 For example, for a new patient, the patient's LDL-C is 3.07 mmol / L, and the predicted Gensini score contribution value is +0.83, indicating that LDL-C increases the patient's risk of coronary artery stenosis. The scores related to each predictor variable of the patient are added together, and finally the EBM model predicts that the patient's total Gensini score is 58 points. The patient's actual total Gensini score is 47 points, which is close, indicating that the model has a good prediction effect on this patient. Figure 4-4 Medium orange indicates positive contributions, while blue indicates negative contributions.

[0143] Step 4: LDL-C risk stratification

[0144] By training on patient data, the EBM model automatically identified the nonlinear change trend of LDL-C on the risk of coronary artery stenosis, so it can accurately select thresholds to stratify LDL-C risk. In this example, the stratification results of LDL-C levels are as follows:

[0145] 1) Low-risk group: LDL-C < 1.9 mmol / L. This group of patients has a lower Gensini score and a lower risk of coronary artery stenosis. When LDL-C is 1.9 mmol / L, the predicted Gensini score reaches zero, indicating that the contribution to the risk of coronary artery disease changes from negative to positive. According to the model prediction, when the LDL-C level is lower than 1.9 mmol / L, the risk of coronary artery stenosis is relatively low, and the prediction results output by the model are highly consistent with clinical observations.

[0146] 2) Medium-risk group: 1.9 mmol / L ≤ LDL-C ≤ 2.7 mmol / L. Within this range, the nonlinear relationship between LDL-C level and coronary artery stenosis is apparent, and the model-predicted Gensini score gradually increases. Although this group of patients is at medium risk, early intervention can significantly reduce the incidence of coronary artery stenosis.

[0147] 3) High-risk group: LDL-C > 2.7 mmol / L. At this time, the LDL-C level is significantly increased, which is significantly positively correlated with the risk of coronary artery stenosis. The model predicts that the Gensini score of patients in this group is higher, indicating that these patients may have more severe coronary artery stenosis and require immediate clinical intervention.

[0148] Step 5: Model Performance Evaluation

[0149] This embodiment can comprehensively evaluate the performance of the EBM model through the following indicators:

[0150] Root mean square error (RMSE): The prediction error of the model was evaluated by the root mean square error (RMSE). On the test set, the RMSE of the EBM model was 47.60, indicating that the model had a high accuracy in predicting the degree of coronary artery stenosis.

[0151] By applying the whole set of methods provided in this embodiment, the following technical effects are achieved:

[0152] 1) Capturing nonlinear relationships: Compared with traditional linear regression models, the EBM model can accurately capture the nonlinear relationship between LDL-C and coronary artery stenosis. Traditional linear models assume that the relationship between variables is linear, while the present invention can better reflect the actual impact of LDL-C on coronary artery stenosis at different levels through the application of the EBM model.

[0153] 2) Personalized risk assessment: This invention realizes personalized LDL-C risk stratification by combining other clinical characteristics of patients (such as age, gender, hypertension, diabetes, etc.). This enables clinicians to provide more accurate treatment recommendations based on the specific conditions of patients, especially in the management of patients with coronary heart disease and acute myocardial infarction.

[0154] 3) High interpretability: The interpretability of the EBM model makes it more clinically applicable than "black box" models such as deep learning. Doctors can clearly see the specific contribution of each predictive factor to the patient's coronary artery stenosis risk through the model's explanatory output, which greatly increases the model's potential for application in actual clinical decision-making.

[0155] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for determining the severity of coronary stenosis. Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0156] like Figure 5As shown, the device 500 for determining the severity of coronary stenosis in this embodiment may include: an actual blood lipid parameter acquisition unit 501, a clinical feature acquisition unit 502, an actual parameter input unit 503, a model output result receiving unit 504, and an actual coronary stenosis severity determination unit 505. Among them, the actual blood lipid parameter acquisition unit 501 is configured to obtain the actual blood lipid parameters of the target patient including low-density lipoprotein cholesterol; the clinical feature acquisition unit 502 is configured to obtain the clinical features of the target patient including identity information and previous medical history; the actual parameter input unit 503 is configured to input the actual blood lipid parameters and clinical features as input information into a pre-trained coronary stenosis severity prediction model; wherein the coronary stenosis severity prediction model is trained using training samples formed by sample inputs consisting of actual blood lipid parameters and clinical features of historical patients and sample outputs consisting of corresponding true coronary stenosis scores, and the coronary stenosis severity prediction model adopts the model framework of an interpretable enhancement machine; the model output result receiving unit 504 is configured to receive the actual coronary stenosis score and score explanation information output by the coronary stenosis severity prediction model; the actual coronary stenosis severity determination unit 505 is configured to determine the actual coronary stenosis severity of the target patient based on the actual coronary stenosis score and score explanation information.

[0157] In this embodiment, in the apparatus for determining the severity of coronary stenosis 500, the specific processing of the actual blood lipid parameter acquisition unit 501, the clinical feature acquisition unit 502, the actual parameter input unit 503, the model output result receiving unit 504, and the actual coronary stenosis severity determination unit 505 and the technical effects thereof can be referred to in detail. Figure 1 The relevant descriptions of steps 101-105 in the corresponding embodiment are not repeated here.

[0158] In some other optional implementations of this embodiment, the actual blood lipid parameters include at least low-density lipoprotein cholesterol and at least one of high-density lipoprotein cholesterol, total cholesterol, and triglycerides.

[0159] In some other optional implementations of this embodiment, the identity information includes: gender information, age information, and past medical history includes: whether there is a history of any of hypertension, diabetes, kidney disease, and stroke.

[0160] In some other optional implementations of this embodiment, before using the sample input and the sample output to form a training sample, the method further includes:

[0161] A missing value interpolation unit is configured to interpolate missing values ​​in sample input and sample output;

[0162] The outlier processing unit is configured to remove or replace outliers in sample input and sample output, so that the data after interpolation and removal or replacement constitutes training samples.

[0163] In some other optional implementations of this embodiment, the missing value interpolation unit is further configured to:

[0164] For continuous variables with missing values ​​in sample input and sample output, median interpolation method is used for interpolation;

[0165] For categorical variables with missing values ​​in sample input and sample output, the majority interpolation method is used for interpolation.

[0166] In some other optional implementations of this embodiment, the score explanation information includes:

[0167] At least one of the following: variable importance analysis results, global explanation information, interaction analysis information, and local explanation information;

[0168] Among them, the results of variable importance analysis are used to present the independent influence of different types of input variables on the actual coronary stenosis score; the global explanatory information is used to present the actual corresponding relationship between the variable value of each input variable and the coronary stenosis score; the interaction analysis is used to present the combined influence of different types of input variables on the actual coronary stenosis score when they are combined with each other; the local explanatory information is used to present the contribution polarity of the variable value of each input variable to the actual coronary stenosis score for the target patient, and the contribution polarity includes positive contribution of positive correlation and negative contribution of negative correlation.

[0169] In some other optional implementations of this embodiment, the score explanation information also includes: the risk level to which the actual coronary stenosis score belongs; wherein the risk threshold for dividing the risk level is determined based on the coronary stenosis severity prediction model by using training samples.

[0170] This embodiment exists as an apparatus embodiment corresponding to the above-mentioned method embodiment. The apparatus for determining the severity of coronary artery stenosis provided by this embodiment adopts an interpretable enhancement machine model that extends the regression tree-based enhancement method on the generalized additive model as a model framework, and adopts blood lipid parameters including low-density lipoprotein cholesterol and clinical characteristics including identity information and past medical history as input variables, so that multiple types of influencing features can be used to jointly predict the accurate relationship between the severity of coronary artery stenosis. Moreover, since the above-mentioned multivariate historical parameters are used as training samples, the characteristics of the interpretable enhancement machine model can provide clear feature explanations and interaction effects between variables while ensuring prediction accuracy, so that the model can also output explanatory information on how to obtain the results, thereby enhancing doctors' recognition of the output results and better assisting doctors in making more accurate diagnoses.

[0171] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can implement the method for determining the severity of coronary stenosis described in any of the above embodiments when executing.

[0172] According to an embodiment of the present disclosure, the present disclosure further provides a readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to implement the method for determining the severity of coronary stenosis described in any of the above embodiments when executed.

[0173] According to an embodiment of the present disclosure, the present disclosure further provides a computer program product, which, when executed by a processor, can implement the steps of the method for determining the severity of coronary stenosis described in any of the above embodiments.

[0174] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0175] like Figure 6As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0176] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0177] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as a method for determining the severity of coronary stenosis. For example, in some embodiments, the method for determining the severity of coronary stenosis may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the method for determining the severity of coronary stenosis described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to execute the method for determining the severity of coronary stenosis in any other appropriate manner (eg, by means of firmware).

[0178] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0179] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0180] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0181] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0182] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0183] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private server (VPS) services.

[0184] According to the technical solution of the embodiment of the present disclosure, by adopting an interpretable booster model that extends the regression tree-based boosting method on the generalized additive model as the model framework, and adopting blood lipid parameters including low-density lipoprotein cholesterol and clinical characteristics including identity information and past medical history as input variables, multiple types of influencing features can be used to jointly predict the accurate relationship between the severity of coronary artery stenosis, and because the above-mentioned multivariate historical parameters are used as training samples, the characteristics of the interpretable booster model can also provide clear feature explanations and interaction effects between variables while ensuring prediction accuracy, so that the model can also output explanatory information on how to obtain the results, thereby enhancing doctors' recognition of the output results and better assisting doctors in making more accurate diagnoses.

[0185] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0186] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining the severity of coronary artery stenosis, characterized in that: include: Obtain the actual blood lipid parameters including low-density lipoprotein cholesterol of the target patient; Obtaining clinical characteristics of the target patient including identification information and medical history; The actual blood lipid parameters and the clinical characteristics are used as input information and input into a pre-trained coronary stenosis severity prediction model; wherein the coronary stenosis severity prediction model is trained using training samples formed by sample inputs of actual blood lipid parameters and clinical characteristics of historical patients and sample outputs of corresponding true coronary stenosis scores, and the coronary stenosis severity prediction model adopts a model framework of an interpretable enhancement machine; wherein the construction method of the model framework of the interpretable enhancement machine includes: A regression interpretable boosting machine model was created. The interpretable boosting machine model used a regression tree to perform step-by-step fitting for each input variable. In each iteration, the model continuously updated the additive structure of each input variable by minimizing the prediction error to capture the nonlinear relationship between low-density lipoprotein cholesterol and coronary artery stenosis. During the model training process, the interpretable boosting machine model automatically identified the interactive effects of low-density lipoprotein cholesterol and other input variables, including age, gender, and hypertension, and found that low-density lipoprotein cholesterol had different effects on coronary artery stenosis in patients of different age groups. receiving an actual coronary artery stenosis score and score interpretation information output by the coronary artery stenosis severity prediction model; Based on the actual coronary artery stenosis score and the score interpretation information, the actual severity of coronary artery stenosis of the target patient is determined.

2. The method according to claim 1, characterized in that The actual blood lipid parameters include at least low-density lipoprotein cholesterol, and at least one of high-density lipoprotein cholesterol, total cholesterol, and triglycerides.

3. The method according to claim 1, characterized in that The identity information includes: gender information, age information, and the past medical history includes: whether there is a history of any of hypertension, diabetes, kidney disease, and stroke.

4. The method according to claim 1, characterized in that: Before using the sample input and the sample output to form the training sample, the method further includes: Performing interpolation processing on missing values ​​in the sample input and the sample output; The abnormal values ​​in the sample input and the sample output are eliminated or replaced, so that the data after the interpolation and elimination or replacement processes constitute the training samples.

5. The method according to claim 4, characterized in that The interpolation process for missing values ​​in the sample input and the sample output includes: For continuous variables with missing values ​​in the sample input and the sample output, interpolation is performed using the median interpolation method; For categorical variables with missing values ​​in the sample input and the sample output, the majority interpolation method is used to perform interpolation processing.

6. The method according to any one of claims 1 to 5, characterized in that: The score explanation information includes: At least one of the following: variable importance analysis results, global explanation information, interaction analysis information, and local explanation information; Among them, the variable importance analysis results are used to present the independent influence of different types of input variables on the actual coronary stenosis score; the global explanatory information is used to present the actual corresponding relationship between the variable value of each input variable and the coronary stenosis score; the interactive analysis is used to present the combined influence of different types of input variables on the actual coronary stenosis score when combined with each other; the local explanatory information is used to present the contribution polarity of the variable value of each input variable to the actual coronary stenosis score for the target patient, and the contribution polarity includes positive contribution of positive correlation and negative contribution of negative correlation.

7. The method according to claim 6, characterized in that The score explanation information also includes: the risk level to which the actual coronary stenosis score belongs; wherein the risk threshold for dividing the risk level is determined based on the coronary stenosis severity prediction model by using the training samples.

8. A device for determining the severity of coronary artery stenosis, characterized in that: include: an actual blood lipid parameter acquisition unit, configured to acquire actual blood lipid parameters including low-density lipoprotein cholesterol of a target patient; A clinical feature acquisition unit, configured to acquire clinical features of the target patient including identity information and past medical history; The actual parameter input unit is configured to input the actual blood lipid parameters and the clinical characteristics as input information into a pre-trained coronary stenosis severity prediction model; wherein the coronary stenosis severity prediction model is trained using training samples formed by sample inputs of actual blood lipid parameters and clinical characteristics of historical patients and sample outputs of corresponding true coronary stenosis scores, and the coronary stenosis severity prediction model adopts a model framework of an interpretable enhancement machine; wherein the construction method of the model framework of the interpretable enhancement machine includes: A regression interpretable boosting machine model was created. The interpretable boosting machine model used a regression tree to perform step-by-step fitting for each input variable. In each iteration, the model continuously updated the additive structure of each input variable by minimizing the prediction error to capture the nonlinear relationship between low-density lipoprotein cholesterol and coronary artery stenosis. During the model training process, the interpretable boosting machine model automatically identified the interactive effects of low-density lipoprotein cholesterol and other input variables, including age, gender, and hypertension, and found that low-density lipoprotein cholesterol had different effects on coronary artery stenosis in patients of different age groups. a model output result receiving unit, configured to receive an actual coronary artery stenosis score and score interpretation information output by the coronary artery stenosis severity prediction model; The actual coronary artery stenosis severity determination unit is configured to determine the actual coronary artery stenosis severity of the target patient based on the actual coronary artery stenosis score and the score interpretation information.

9. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for determining the severity of coronary stenosis according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method for determining the severity of coronary artery stenosis according to any one of claims 1 to 7.

Images